On Wednesday 25 January 2006 19:42, Markus F.X.J. Oberhumer wrote:
> Hello Andreas,
>
> Andreas Gruenbacher wrote:
> > On Wednesday 25 January 2006 19:05, Markus F.X.J. Oberhumer wrote:
> >> Indeed, you are right. Still it seems that the optimization might make
> >> sense in the second ("what_restore") case.
> >
> > No, the backup file may also only have a link count of one. There's
> > another weakness in the code though: the backup file is only unlinked
> > after ensure_nolinks was called, at which point a copy has already been
> > created. I'm fixing that as well now.
>
> I will have to look at the code once more again, but I still think there is
> an unneeded link/copy/unlink/rename cycle happening in some cases.
That's indeed the case, but link/unlink/rename are cheap compared to copying,
so I think it doesn't make sense to special case this in the code.
> BTW, please also apply the small patch below - checking for ENOSYS is a
> good idea esp. now that we're beginning to see a number of interesting FUSE
> filesystems.
Makes sense, yes.
Andreas
_______________________________________________
Quilt-dev mailing list
[email protected]
http://lists.nongnu.org/mailman/listinfo/quilt-dev
